Evidence of Transference and Emergence in the Interlanguage

By

Philip J. Monahan University of Florida

May 2001

1

Acknowledgements

The completion of this work could not have been possible without the help of two people, in particular, among many. The first is Caroline Wiltshire, whose comments, insights, support and motivation aided immeasurably in the conception and completion of this paper. Her critiques and positive criticism throughout the entire research process was truly invaluable. Furthermore, as a truly wonderful educator, I acquired a true passion for linguistic, and more specifically, phonological inquiry from her teachings and lectures. It is rare for someone to find what they really love, and for this I feel truly lucky. This type of inspiration is truly priceless. The second individual is Patricia McCord, whose unyielding devotion, love and support for me is something that kept me going and kept me dedicated day in and day out. Others that need to be mentioned here for their support are, very importantly, my family, who has always allowed me to make my own decisions, while at the same time continued to support me, even though I still believe that they are not really sure what linguistics is. Also, many people associated with the linguistics program at the University of Florida have been involved in helping with the project. They include Ratree Wayland, whose acoustic equipment was used, Eric Potsdam, Jodi Bray, Russell Moon, Jodi Nelms, and Paul Kotey. Their aid can probably never be repaid, so thank you. Finally, this research project was funded by the University of Florida Undergraduate University Scholars Program for the academic year 2000-2001. This is truly a wonderful program that allows undergraduate students the opportunity to partake in very satisfying and fulfilling original research. Thanks, also, to anyone I may have overlooked, all of your help has been truly appreciated.

2

Evidence of Transference and Emergence in the Interlanguage

Philip J. Monahan University of Florida pmonahan@ufl.edu

“A journey of a thousand miles begins with a single step” Chinese Proverb

0.Introduction………………………………………………………..…………………3 1.Background…………………………… …………...……………………………......5 2 The Syllable Structures of Brazilian Portuguese and English……………...………..7 3 The Vowel System of Brazilian Portuguese………………………………………...13 4.Methodology………………………………………………………………...………17 5 Phonetic and Phonological Analysis………………………………………………...19 6 OT: Constraint Ranking from L1 to L2……………………………………………..27 6.1 OT: Introduction and Constraints…………………………………………28 6.2 Brazilian Portuguese Ranking…………………………………………….28 6.3 Interlanguage Ranking………………………………………………… …37 7 Conclusion………………………………………………………………..…………47 8 References………………………………………………………………………...…49

3

Evidence of Transference and Emergence in the Interlanguage 0. Introduction

Second language learners have a tendency to transfer phonological constraints and

pronunciations from their first language (L1) into their second language (L2) (Gass and

Selinker 1994). Optimality Theory (Prince & Smolensky 1993) captures the phonology of

the L1 using a system of ranked constraints. The ranking of constraints provides us with a

detailed understanding of the phonological processes that occur in the language. For

example, which constraints are higher ranked or more prominent within that language, as

well as which constraints the language is willing to violate in order to satisfy the higher

ranked constraints is inherent within the theory. The goal of this paper is to rank the

constraints of Brazilian Portuguese (henceforth BP) syllable structure, adopting an

Optimality Theory approach, and determine if native speakers of BP transfer the ranking

into their L2 English.

First, the paper gives a brief overview of syllable structure, and secondly, it discusses

aspects of the phonological system of BP (i.e. regressive nasalization, the vowel system,

and lateral/vowel gliding), including constraints on the syllable structure (onset and

rhyme) as compared to those in English. Next, the methodology of the data collection is

presented. The data collected were English sentences read by native speakers of BP that

contained phonological constructions found in English but not in BP. Then, a phonetic

and phonological analysis has been done on the collected data and is presented in §5.

4

Next, the constraints on BP syllable structure are ranked, and finally that ranking is

compared and contrasted to the L2 English ranking, as well as L1 English.

It was found that the constraint on the coda in BP was transferred high in the

interlanguage ranking, and thus processes like regressive assimilation of nasality/nasal

deletion and lateral gliding were apparent in the interlanguage phonology. Furthermore,

an constraint that captured the unmarked structure deviant from both BP and English

emerged in the interlanguage. However, the data did not support the hypothesis that

epenthesis would occur when a native-speaker of BP was confronted with complex

syllable margins, as in English.

Below are the consonant inventories, to the best of my knowledge of BP and

English. These tables contain only the phonemic segments apparent in the languages.

0.1 The Consonant Inventory of Brazilian Portuguese Place

manner Labial

Labial Dental

Inter-dental

Alveolar Alveo-palatal

palatal velar glottal Labial-velar

Stop p b t d k g

fricative f v s z S Z

nasal m n ø

affricate tS

rhotic r R

approx. j w

lateral l ´

The consonant inventory of BP contains / ø ´ r /, which do not appear in the

consonant inventory of English.

5

0.2 The Consonant Inventory of English Place

manner Labial

Labio-dental

Inter-dental

Alveolar Alveo-palatal

Palatal Velar Glottal Labio-velar

stop p b t d k g

fricative f v T D s z S Z h

nasal m n N

affricate tS dZ

rhotic ¨

approx. j w

lateral l

The consonant inventory of English contains / T D N h dZ ¨ /, which do not

appear in the consonant inventory of BP.

1.1 The Syllable

The syllable is phonologically composed of two parts, the onset and the rhyme, which in

turn consists of a nucleus and a coda. The nucleus of the syllable is generally a vowel,

except in marked cases, such as Berber (a language spoken in Morocco), where

consonants are sometimes used in the nuclear position, such as the word txdmt 'to gather

wood' (Archangeli 1997). As for the onset and coda, languages have constraints on which

segments are allowed to occur in consonant clusters in each position. For instance,

English, as we will discover, has more permissible consonant clusters than BP in both the

onset and the coda. English allows up to three consonants to fill the onset and two in the

coda. A well-known example is the English word strengths, when syllabified:

6

(1) Maximal Syllable in English ó / \ / R / / \ O N C //\ / //\

s t r E N T (s)1

1.2 The Sonority Hierarchy

English consonant clusters are also restricted with regards to the sonority hierarchy. The

sonority hierarchy, ranking segments from least to most sonorous, is as follows: voiceless

obstruents > voiced obstruents > nasals > liquids > glides > vowels. In English, as in

other languages, permissible syllable constructions tend to follow this hierarchy going

from least to most and back to least sonorous (in closed syllables) within the syllable. For

example, again using the English word strengths, the syllable begins with two voiceless

obstruents (the fricative will always precede the stop mono-morphemically in English)

which are least sonorous on the hierarchy. The liquid [r] follows, which is higher than the

voiceless obstruents but lower than the vowels on the hierarchy. Next, is the vowel [E],

the highest on the sonority hierarchy, before falling back down the hierarchy, from the

nasal to the voiceless obstruents. This applies in BP syllable structure as well, but only in

the onset, because that is the only part of the syllable where consonant clusters can occur.

Example (2) shows the sonority hierarchy applied to the English words grump and smart.

(2) _ _

_ # - _ _ Ü - _

g r à m p s m a r t

1 In some accounts of syllable structure, the third consonant in the coda in English is syllabified into the word appendix and not the syllable coda (Spencer 1996).

7

In the example, the lines represent the levels of sonority for the individual

segments, where the highest lines are most sonorous and lowest are least. Notice the

gradual rise and fall in sonority throughout the word. This holds true for all words in

English and BP. Next, a similar representation will be given for the BP lexical items tres

‘three’ and flor ‘flower’ to show that BP also follows the sonority hierarchy in terms of

which consonants are permissible in clusters.

(3) _ _

_ - _ _ Ü _ t r e s f l o r 2.0 The Syllable Structure of Brazilian Portuguese and English

The following section will present the basic workings of Brazilian Portuguese phonology,

much of which will be necessary for our constraint-based analysis that follows in § 7.

Also, in this section the onset and coda of English will be discussed to show how the

languages are different, as well as how they are similar in relation to the permissible

clusters and the sonority hierarchy.

2.1 The Onset in Brazilian Portuguese

The most typical consonantal cluster in BP is plosive + approximate, and except for a few

fricative + liquid cases, all other sequences are impossible in Portuguese, regardless of

whether or not they follow the sonority hierarchy (Mateus and d’Andrade 2000). In the

onset there tends to be no debate as to how many consonants can occur and which can

occur sequentially. All Portuguese consonants occur syllable-initially (with the exception

of the alveolar trill [r]), but prenuclear consonant clusters are restricted. The only possible

8

combination is an obstruent /p b t d k g f v/ followed by an approximant, and there are

certain obstruents that cannot occur with the liquid in the onset. For example, [dl] does

not occur except across morpheme boundaries in borrowed lexical items, which means

that it is not syllabified in the onset. The clusters [vr] and [tl] never occur, and [vl] is a

rare case occurring in borrowed words only, such as Vladimir and cross syllables, as in

the brand name Revlon (Azevedo 1981), which also means that it is not syllabified in the

onset.

In BP, we find that clusters made of segments with identical sonority values (i.e.

two stops or two fricatives *[pt], *[sf], etc.) do not occur, nor do plosive + fricative (i.e.

*[tf], *[ts], etc.) or fricative + nasal (i.e. *[sn], *[fm], etc.) clusters. Furthermore, BP does

not allow fricative + rhotic, except in a few cases such as, [fR]io ‘cold’, pala[vR]a ‘word’

and re[fR]escar ‘to refresh’, nor does it allow fricative + lateral clusters, except in a few

such cases as [fl]or ‘flower’ and a[fl]orar ‘to emerge’ (Mateus and d’Andrade 2000).

One more comment regarding onset clusters needs to be made. In Portuguese,

non-permissible onset clusters appear in the input forms (underlying forms and

orthographically), in such cases as *pn pneu ‘tyre’, *gn gnomo ‘gnome’, and *ps

psicologia ‘psychology’. In these examples we find epenthesis, whereby [i] is inserted to

break up the impermissible cluster. For example, pneu is [pi-ne @w] and gnomo is [gi-

n�@mu] (Mateus and d’Andrade 2000). This shows that when native speakers of

Portuguese are confronted with non-native like clusters they use epenthesis in order to

retain the ideal CV syllable structure.

9

2.2 The Onset in English

The restrictions on the onset in BP are, in general, similar to those on the onset in

English, except that the onset in English is a little more flexible. English, like BP, allows

onsetless syllables and any consonant with the exception of [N] can stand alone in the

onset. English allows two consonants to occupy the onset position of the syllable, where

the first must be an obstruent, followed by an approximant /l r w j/, unless the first

consonant is [s], then nasals, with the exception of [N], voiceless stops and approximants

can occur in the second position of the cluster. English does, however, permit three

consonants to occupy the onset of a syllable, where the first must be /s/, followed by a

voiceless oral stop, followed by an approximant. Furthermore, the consonant clusters in

the onset of both languages must be rising in sonority.

2.3 The Coda in Brazilian Portuguese

The coda in BP undergoes far more restrictions than the onset, and the number of

segments in the coda position of the syllable in BP has been debated extensively in the

literature. Girelli (1988) argues against BP allowing possible CVC type syllable

constructions. According to Girelli, in his X-bar theory approach to phonology,

consonants (namely /s/ in his account) are allowed to follow the nucleus but they do not

occupy the coda position of the rhyme. Instead, the segment is part of the nucleus, and

the only allowable segment in this position is [s]. He also makes note of /n/ regressively

assimilating its nasality onto the preceding vowel, and /l/ gliding and adjoining onto the

nucleus. These issues will be addressed later in the paper (see §2.5 and § 3.3,

respectively).

10

Mateus and d’Andrade (2000) argue that BP does allow the phoneme /s/ to appear

in the coda position of the syllable, and in this position it is phonetically realized as /S/ or

/Z/ depending on the voicing of the initial consonant of the following morpheme.

Azevedo (1981) notes that the segments /r s l n/ are all allowed in the coda

position of BP and that the sequence [ns] is the only consonant cluster found in the coda.

But, it needs to be mentioned that in Azevedo’s analyses the [n] nasalizes onto the

preceding vowel and is deleted while leaving the [s] as the only consonant occupying the

coda position on the surface. For example, campa ‘handbell’ becomes [ka )pa] and interim

becomes [i@teri )].

BP allows [s] and [r] in coda position, however, /l/ and /n/ only occur in the

underlying forms and do not appear in surface representations. This is an important point

because as we will see in the OT analysis, BP wants to mi nimize its coda constituents,

and both the /l/ and /n/ are lost because of it, especially when /ns/ occurs in the

underlying representation. The phonotactics of BP will force regressive nasalization of

the vowel and deletion of the nasal consonant, thereby avoiding consonant clusters in the

coda position of the syllable (see §2.5).

2.4 The Coda in English

The possible consonant clusters permissible in the coda of the syllable in English are far

more flexible than those in BP. For example, English allows up to two consonants in the

coda, and one or two more occurring in the appendix of the syllable, such as in the word

strengths [strENTs] discussed above (see §1.1). This is as opposed to BP, which allows

11

only one consonant in the coda, and it must be an allophone of /s/ or /r/. The consonants

in the coda in both languages must also be falling in sonority.

2.5 Regressive Nasalization and Nasal Deletion in Brazilian Portuguese Wetzels (1997) classifies nasal vowels in BP as being either contrastive or allophonic,

and labels the former “nasal vowels” and the latter “nasalized vowels.” He argues that

nasality is systematically realized on the vowel. Furthermore, nasal vowels in BP occur in

both stressed and unstressed syllables, both word-internally and word-finally. Examples

appear below (from Wetzels 1997):

(4) Nasalized Vowels Surface Representation Underlying Representation Spelling Gloss a. word-internal stressed

[f i )ka] /finka/ finca ‘fixes’

[pe )tSi] /pente/ pente ‘comb’

[ka )pa] /kampa/ campa ‘handbell’ b. pretonic

[u)bi @gu] /umbigo/ umbigo ‘navel’ c. word-final stressed

[kupi)] /kupim/ cupim ‘termite’ d. word-final unstressed

[i@teri )] /interim/ interim ‘interim’ e. nasal consonants in onset position [anuaw] /anual/ anual ‘annual’ [natu] /nato/ nato ‘native’ [mata] /mata/ mata ‘wood, forest’

These examples show that regressive assimilation of nasality is occurring onto the

preceding vowel and that the nasal consonant in the underlying form is being deleted.

This plays a role in the construction of the coda in BP, in terms of which segments are

allowed. The syllable of BP allows for only three positions in the rhyme, and the non-

nuclear position tolerates only /s l r n/ in the underlying form and only [s] and [r] on the

12

surface. If the nucleus of the rhyme is nasalized due to regressive assimilation, and the

nasal consonant is deleted, then no other consonant can occupy the coda position except

[s] (Wetzels 1997).2 To further explain the repercussions nasal assimilation and the

deletion of the nasal consonant have on the syllable structure of BP, we will look across

morpheme boundaries. When we regressively nasalize a vowel, the nasal consonant is

deleted, as was mentioned above, but its phonological properties, namely its position in

the coda, remain in the underlying representation. For instance, in BP a nasal followed by

/l/ violates the phonotactics of BP consonantal clusters, and is therefore disallowed. The

same is true across morpheme boundaries. Therefore we will never find a nasalized

vowel followed by a /l/. In fact, the only place we find /C/ + /l/ cross-morphemically is

when /C/ is an allophone of /r/ (Wetzels 1997).

Nasal vowels in BP are contrastive when followed by a non-nasal consonant, as

well as when they occur word-finally (Reider 1981). For example, [ka )pa] ‘tombstone’

and [ka @pa] ‘cloak’ and [la )] ‘wool’ and [la@] ‘there’. However, as was previously

mentioned, this is not the case when the vowel occurs before a nasal consonant.

Reider (1981) argues that traditionally nasalized vowels in BP, vowels occurring

before a nasal consonant, are becoming denasalized in non-stressed positions in northern

and central regions of Brazil.

(5) Traditionally Northern/Central dialects of BP

[ba )na )na] [bana )na] ‘banana’

[a )ni )ma )w )] [anima @w] ‘animal’

2 It should be noted that regressive assimilation of nasality does not always occur, and for a more detailed analysis, see Wetzels 1997.

13

Participants from these regions were not used in the this research project because

of this reason.

3.0 The Vowel System of Brazilian Portuguese

3.1 The Inventory

Portuguese has seven oral vowels / i u e o E � a / (Wetzels 1997), and Mateus and

d’Andrade (2000) argue for an eighth /�/, which are found in stressed position only. In

unstressed position the vowels undergo a variety of phonological processes, which will

be addressed later in the paper (see §3.2 - 3.4). The vowels tend to follow the same

features (height, backness, and rounding) as they do in English, with /i/ and /u/ being

[+high] and [+ATR] and so forth.3

3.2 Diphthongization In BP vowels in stressed positions have a strong tendency to become diphthongs. In the

Cario@ca dialect these vowels typically lengthen and add a [j] or [w]. This is also seen in

the speech of young children in the Carioca dialect, for instance, [pajs] paz ‘peace’ and

3 There has been argumentation over the actual placement of /a/ in the vowel chart. Harris (1974) argues that Portuguese has three vowel heights, which are distinguished by tongue height. Redenbarger (1981), who proposes that Portuguese has the vowel /æ/ in its inventory, also claims that Portuguese has three vowel heights, but two layers of mid vowels, where the high vowels are realized by tongue height, and the mid vowels, namely /a/ and /æ/ in his analysis, by the feature constricted pharynx (known in his account as [+cp]). This feature is reliant on articulatory and acoustic properties, rather than phonological representations. It was created by Redenbarger in order to allow the distinctive feature-based system of phonology proposed by Chomsky and Halle (1968) in the Sound Pattern of English to fully account for the vowel /a/ in Portuguese. According to his analysis, SPE, within the features proposed, could not distinguish between the Portuguese vowels /a/ and /æ/. He concluded that by adding the distinctive feature [+cp] it would eliminate the phonological ambiguities between /a/ and /æ/ when drawing up the framework of a vowel chart which incorporates all the vowels into their correct position, both acoustically and articulatorally, and for these positions to hold in phonological analysis.

14

[‘n�ys] no@s ‘us.’ BP nasal vowels /e) o)/, in many dialects, become diphthongs in primary

stress and word finally, such as [bo)w )] bom ‘good’ and [ale)j )] alem ‘besides’ (Major 1985).

3.3 Liquid and Vowel Gliding and its Effects on Syllabification It can be predicted that high vowels become glides when occurring immediately after a

stressed vowel and that VG diphthongs occur both tonically and pre-tonically (Giangola

1997).

(7) Vowel Gliding V [+high] à G / V [+stress] ___

a) ai à [a@j] ‘woe’

b) sarau à [sa.ra@w] ‘evening party’ VG Diphthongs a) eu à [ew] ‘I’

b) euforia à [ew.fo.ri@.Ã] ‘euphoria’

Generally when an intervocalic glide occurs, VGV, the glide is traditionally

syllabified in the coda. Many dialects of BP will epenthesize a glide between the stressed

vowel and the [s] if occurring word-finally (from Giangola 1997).

(8) a) [lujs] luz ‘light’

b) [dEjs] dez ‘ten’

c) [hapa@js] rapaz ‘guy’

Post-tonically, [l] undergoes gliding when syllabified in the coda, and remains a

lateral when syllabified in the onset (Giangola 1997).

(9)

a) [fa.siw] fa@cil ‘easy’

b) [fa.si.li.mU] faci@limo ‘easy (superlative)’

15

The affixation in this case causes the morpheme to be syllabified differently. In a) we

see that the liquid is syllabified in the coda, and therefore undergoes gliding. However, in

b), the addition of the superlative morpheme causes the liquid to be syllabified in the

onset, where no feature changes take place (Giangola 1997).

The glides discussed so far would all be considered off-glides, as opposed to on-

glides, which will be discussed next. In BP on-glides also occur, and they are syllabified

in the onset of the syllable (from Giangola 1997).

(10)

a) [la@.bjU] labio ‘lip’

b) [la.bi@.aw] labial ‘labial’

aa) [ta @.bwÃ] ta@bua ‘board’

bb) [ta.bu.a@.dÃ] tabuada ‘multiplication table’

Examples a) and aa) in (10) exemplify the point that the on-glide is syllabified in

the onset of the syllable in BP. However, when an additional morpheme is added, stress is

realized on a different syllable, and here we find that the glides [j] and [w] are realized as

their vowel counterparts [i] and [u], respectively. Glides in the nucleus position appear as

vowels.

3.4 A Template for the Syllable Structure of BP and OT in SLA From the phonological issues addressed so far, a surface syllable template will be

formulated for Brazilian Portuguese.

(11) ó / \ / R / / \ O N C /\ / /\

(C)(L/G) V (G)(C)

16

From this template of the syllable structure of BP, we can see that any consonant

occurring in Portuguese is allowed in the onset. But as noted before, the permissible

obstruents occupying the consonant of consonant/approximant cluster are limited. Next,

we find the vowel (diphthongs are also present as indicated by the V(G) representation

dominated by the nucleus (N) node) occupying the nucleus position, and in the coda we

find only one position, which can only be occupied by some variant of /r/ or /s/ and not

both. Remember that /l/ and /n/ are permissible coda consonants in the underlying form,

but they are glided or force regressive nasalization on the previous vowel and deleted,

respectively, in the surface form. Notice also that the only obligatory component of the

syllable is the vowel occupying the nucleus position in the syllable.

Before moving onto the methodology and analysis sections of the paper, it is

important to briefly discuss second language acquisition and interlanguage phonology as

it has been treated in OT. One of the better known studies is Broselow, et al. (1998),

where the interlanguage production of obstruents in the coda position by native speakers

of Mandarin Chinese learning English was analyzed using an OT approach. What was

found was quite interesting in the sense that the language learner preferred unmarked

forms that are not native to either Chinese or English. For example, if the form was

monosyllabic, the speaker would epenthesize to utter the form as disyllabic, and if the

form was disyllabic, the speaker would delete the final obstruent in the coda position.

Other times, the speaker would devoice the obstruent, as voiceless obstruents are less

marked than voiced ones. Essentially the study showed that disyllabic forms, less marked

than any other word construction, and voiceless obstruents were preferred in the

interlanguage phonology. These processes do not occur in English or Chinese, which

17

illustrates an interlanguage phonology that can deviate from both the native and target

languages. According to Broselow et al., these forms result from low ranking constraints

that are masked or hidden in the native language becoming higher ranked because of the

richness of the base of the target language. The interlanguage grammar thus differs from

the native-like grammar. This reranking of constraints is responsible for differences in the

phonology of the interlanguage from both the native or target language, and as the

language learner continues to learn the target language, the constraints continue to be

reranked, moving closer to the ranking of the target language.

4.0 Methodology4

This study was designed to determine the interlanguage constraint ranking of native

speaker’s of BP learning English. The purpose was to determine if this ranking matched

the native language ranking, the target language ranking, or if unmarked structures that

deviated from both the native and target languages emerged in the interlanguage

phonology, as argued by Broselow et al.

4.1 Participants Five native speakers of Brazilian Portuguese pooled from the University of Florida and

Santa Fe Community College student bodies participated in the project. Their ages

ranged from 21 to 28 years old, and they have resided in the United States from four

months to three years. Most of them studied English while in Brazil and have studied

English for more than a year.

4 The sentences containing the target words are found in appendix A.

18

4.2 Procedure The participants were asked to read 67 English sentences at three different rates of

speech, careful (slow), casual (normal) and fast (quickly). These sentences contained a

target word embedded within them, isolated by vowels on either side. Example (12)

provides the template of the sentence construction used when collecting the data, where

the X represents the position of the target word in the sentence.

(12) I will say X again.

The target words were chosen because they contained either a consonant cluster

non-existent in BP or a segment or phoneme not present in the BP segmental inventory.

For example, as was previously mentioned, BP does not allow complex consonant

clusters in the coda of the syllable, but English does, so a word like twelfths was chosen

to be a part of the data set. The choice for these words was based on the idea that if a

certain construction or phone is not present in a particular language, then when the native

speaker learns a second language that may contain those constructions or phones, the

speaker will introduce traits of their native language into their second language. For

example, BP does not allow complex consonant clusters in the coda of the syllable,

therefore when confronted with a word like twelfths we would expect the second

language learner to epenthesize and break up the “foreign” consonant cluster.

The participants were then asked to read two paragraphs, which contained a large

percentage of the selected words in the sentences. In the paragraphs, however, the target

words were placed before and after words with consonants and consonant clusters in the

edges. This was done to vary the environments in which the target words occurred. For

instance, in the sentences the target words were located between a word ending in a

19

vowel, say, and a word beginning with a vowel, again. In the paragraph, the goal was to

place these target words in different environments. The goal was to determine how, for

example, the native speaker would adjust if the target word began with a consonant

cluster and followed a word that ended in a consonant cluster.

The participants were recorded in a sound proof booth located in the basement of

the Department of Psychology at the University of Florida. The data was recorded onto a

digital audiotape (DAT) using a SONY DAT recorder. The data was then burned onto a

compact disc and broken up using Cool Edit from Syntrillium. Finally, the data was

analyzed on a spectrogram Wave Surfer 095 from Scriptics Corp., to determine processes

that could not be distinctively deciphered by just listening to the tape. Such processes as

regressive nasalization and vowel deletion, lateral gliding in the coda, and vowel

epenthesis to break up non-native like clusters, among others, were acoustically analyzed.

5.0 Phonetic & Phonological Analysis5 The aim of this section is to explain the phonetic and phonological findings of the

interlanguage English of native BP speakers, as extracted from the data collected. In the

previous sections such processes as epenthesis to alleviate impermissible consonant

clusters, regressive assimilation of nasality and nasal deletion in the coda position, and

lateral gliding in the coda position have been discussed. The focus of the rest of the paper

is to assess three of these processes in terms of the interlanguage English (§ 5), the

constraint ranking motivating these alternations in BP (§ 6), and finally the comparison of

the interlanguage constraint ranking to English and BP (§ 6). The three processes to be

5 Sample acoustic spectrograms from the analysis are found in appendix B.

20

addressed are epenthesis (§5.1), regressive assimilation of nasality and nasal deletion in

the coda position (§5.2), and lateral gliding in the coda position (§5.3).

5.1 Epenthesis To recall from §2.1, BP epenthesizes a vowel between two underlyingly adjacent

consonants that violate the phonotactics of permissible consonant sequences. For

example, the form gnomo surfaces as [gi.n�.mu] ‘gnome’ with the epenthetic [i] surfacing

to break up the /gn/ cluster. According to the second language acquisition literature,

processes like this should occur in the speaker’s interlanguage when the speaker is

confronted with a consonant cluster not occurring in their native language.

This expectation was tested in the data collection, where non-BP like consonant

clusters found in English were placed in both the onset and coda positions of the syllable

(see §1.1 for a review of the syllable). Complex syllable margins in English words such

as strengths, script, scripts, sixth, six, burst, stow, spay amongst many others (for the

complete list consult Appendix A) were tested in the data collection. The general notion

was that the native speakers of BP would epenthesize at some place in the target word in

order to break up the non-BP consonants clusters presented to them. For instance, when

considering the word stow, the expectation would be epenthesis, similar to Spanish, by

adding an [E], or some other vowel, at the beginning of the morpheme in order to

resyllabify the [s] into the coda of the new syllable, while [t] remains in the onset of the

original syllable. This eliminates the [st] onset cluster not by epenthesizing to break up

the cluster but rather by epenthesizing to breakup the onset into two syllables. So, the

expected utterance of a native speaker of BP pronouncing stow would be [Es.tow],

21

thereby syllabifying the [s] into the coda of the syllable formed around the epenthesized

[E], while leaving [tow] to remain as a well-formed syllable in both English and Brazilian

Portuguese. Continuing this section is an explanation of the acoustics of vowel

epenthesis, as well as what was found in the collected data, where the question “do native

speakers of BP epenthesize to make their English utterances well-formed BP syllables?”

is posed and answered negatively based upon the collected data.

In a spectrogram analysis, an epenthetic vowel has the same acoustic

characteristics as that of a non-epenthesized, or input faithful, vowel depending on the

actual vowel itself. For example, an epenthesized [u] has the same acoustic characteristics

as an input faithful [u], where the formants are similar and differ only, of course, based

on transitions to following segments.

In a very small number of cases the native speakers of BP did epenthesize where

expected; however, in general, the speakers did produce most utterances in a native-like

way, in the sense that epenthesis did not occur. The forms used in the constraint analysis

of epenthesis are, for the most part, input faithful.

The isolated cases illustrated the expected phonological alternations, in terms of

epenthesis, whereby a vowel, [E] when epenthesized at the beginning of the root, and [«]

elsewhere. For example, scripts was produced as [skrI.p«ts] and [Es.krIpts], where in both

cases a vowel was epenthesized to break up at least one non-BP cluster. Another example

22

of this was the word strengths, which in some cases was produced as [strE.Ν«s], whereby

[«] was epenthesized to break up a [Νs] cluster.6

5.2 Regressive Assimilation of Nasality and Nasal Deletion in the Coda Position Previously, regressive assimilation of nasality and nasal deletion was discussed (§2.5) to

illustrate that BP regressively assimilated nasalization onto a vowel from a nasal

consonant in the coda position and deleted the nasal consonant. As before, the same

phonological alternation is expected to occur in the English of native BP speakers.

Within the data set a number of English words which had the nasal in the coda

were tested to determine if native BP speakers transfer this process into their

interlanguage English. For example, the form plant appeared in the data, and it would be

expected to surface as [plæ)t], as the nasalization is assimilated and [n] is deleted. To

illustrate this process more clearly, figure (13) presents a derivational account of this

expected alternation using the English form plant.

(13) A Derivational Account of Regressive Nasality and Nasal Deletion in the Coda Position /plænt/ ⇓ [plæ)nt] Regressive Nasality ⇓ [plæ)t] Nasal Deletion

6 Here, the omission of [Π] is not really relevant as [Νs] itself does not constitute a well-formed consonant cluster in BP and [Π] is not a naturally occurring segment in BP. Deletion is beyond the scope of this paper.

23

This process is indeed evident throughout the data, as it was suspected to be, with

a few non-notable exceptions. Throughout the data, heavy nasalization of vowels was

present and in most cases there was no, or in other cases very little, evidence of the nasal

consonant surfacing.

Acoustically, nasalized vowels are realized by a reduction in the intensity of the

first formant, while the third formant is typically higher than the third formant of the fully

oral vowel. The third formant is rising in lightly nasalized vowels, and in fully nasalized

vowels the third formant is fully high with no rising or falling in frequency (Ladefoged

and Maddieson 1996). This is seen in most of the cases in the data, where the first

formant of the vowel is decreased in amplitude and the third formant is high. This

illustrates that the first step of the process, regressive assimilation of nasality, is evident

in the interlanguage; next, the second step, that of nasal deletion in the coda, will be

addressed.

Acoustically, a nasal consonant, [n] especially, is realized with a first formant

around 280 and a nasal zero at 1780. In other words, a nasal consonant is acoustically

realized with one formant at 280 and nothing at the higher frequencies (Ladefoged and

Maddieson 1996). These formant characteristics are evident throughout the data. In the

native English speaker’s speech, the first formant remained around 280 for the duration

of the nasal articulation; however, in the non-native’s speech, the first formant was

indeed not present and no traces of a nasal articulation were found with the exception of

the nasalization features on the preceding vowel. Therefore it can be concluded that this

constraint, one against nasal consonants surfacing in the coda position, was transferred

24

into the interlanguage. The resulting surface representations of the English forms are

listed below in (14):

(14) Surface Representations of the Nasal Forms a. English b. BP Interlanguage English c. Gloss

[plæ)nt] [plæ)t] ‘plant’

[klæ)n] [klæ)] ‘clan’

[o )w)nz] [o )w)z] ‘owns’

[a )w)ns] [a )w)s] ‘ounce’

The forms in (14a), the English forms, are as they would appear by a native

speaker of English, in which the vowel is nasalized due to the following nasal consonant,

which regressively assimilates its nasality onto the preceding vowel. The nasal consonant

is articulated and if another consonant remains in the coda (or appendix) position then it

too is articulated. The forms in (14b), the BP interlanguage English, illustrate that

nasality is regressively assimilated, however, the difference lies in that the nasal

consonant is deleted. This follows from the phonology of BP.

5.3 Lateral Gliding in the Coda Position The final phonological alternation to be discussed is that of lateral gliding in the coda

position. It has already been discussed that BP is very restrictive in terms of coda position

(§2.3), and in cases where a lateral is present in the underlying form it is glided on the

surface and syllabified into the nucleus of the syllable (§3.4). This is illustrated in (15) (a

recapitulation of 9a and 9b):

(15) Lateral Gliding in the Coda Position Underlying Surface Gloss English Gloss

a. /fa @.sil/ [fa @.siw] facil ‘easy, simple’

b. /fa.si.li.mU/ [fa.si.li.mU] facilimo ‘easy (superlative)’

25

The example in (15) illustrates that when a lateral is syllabified in the coda

position (typically word-finally) it is glided and syllabified into the nucleus. However,

using the same underlying form and suffixing a vowel-initial morpheme, which forces the

lateral into the onset position of a syllable, the segment surfaces as a lateral. This

illustrates that 1), the underlying form is lateral and 2), when that lateral is syllabified in

the coda it is glided, and when it is syllabified in the onset then it surfaces as a lateral.

As in the previous two cases discussed so far, this process is expected to be

transferred into the interlanguage English of native BP speakers, and like nasalization

(§5.2), it is evident that this process is transferred. First, the acoustic characteristics of

lateral gliding will be discussed and finally the surface forms that will be used in the

constraint analysis are introduced.

A lateral in the spectrogram is indicated by antiformants occurring on the higher

vowel formants, indicated by a weakening in amplitude. A glide in spectrogram analysis

is characterized by a change in the formant frequencies of the preceding vowel. For

example, in the form stow, the vowel [o] glides to [w] and this is reflected in the

transitions of the formants. For instance, depending on the vowel and the glide under

inspection here, the formants of the vowels may diverge, converge, or rise or lower in

unison. In English, pure vowels are generally rare, so in almost all cases there is some

formant transition, as there is in BP. However, the difference is evident in the presence of

the lateral and the length of transition. In the English data, the transition from vowel to

glide is relatively short and the lateral is present and indicated by a strong low first

formant, a weaker second formant and a third stronger high frequency formant

(Ladefoged and Maddieson 1996). In many of the cases, in the BP speakers’ data the

26

gliding was perceptually evident. Ladefoged and Maddieson (1996) argue that, in fact, in

BP the raising of the back of the tongue creates a segment similar to [U] and this segment

merges with the lateral to become a vowel or semi-vowel.7

The forms used in the constraint analysis are presented below. The schema in (16)

illustrates the differences between the English and BP interlanguage English utterances.

(16) Lateral Gliding in the Coda Position English Interlanguage English Gloss

a. [t�l] [t�w] ‘tall’

b. [�l] [�w] ‘all’

The examples in (16) illustrate the differences in the pronunciation of the English

forms tall and all. The gliding in the English forms is not extremely evident and is

therefore not included in the transcriptions. The interlanguage forms are lacking the [l];

however, they do show significant gliding and it is therefore evident in the transcription.

5.4 Conclusion of Analysis

This concludes the section on the phonetic and phonological analysis of the collected

data. Three topics of particular importance to the constraint analysis below were

discussed, that of epenthesis (§5.1), regressive assimilation of nasality and nasal deletion

(§5.2), and lateral gliding in the coda position (§5.3). It was evident from the data that

both nasalization/deletion and lateral gliding were wide spread; however, epenthesis was

not. Based on this data, regressive assimilation of nasality, nasal deletion, and lateral

gliding in the coda position seem to be transferred into the interlanguage of the BP native

7 However, because of articulatory features, the lateral will continue to remain as classified as such. This point is not directly of interest, as their concern resides more in classification than acoustic or phonological for this particular point.

27

speaker’s English. Epenthesis did not always occur in the forms it was expected to, or at

least not enough to generalize as in the other two processes of interest. Such clusters as

[rst], [str] and approximant plus obstruent clusters posed little or no difficulty to the

native BP speakers. There are some possible explanations that will be discussed in the

conclusion of the paper.

6.0 OT: Constraint Ranking from L1 to L2 This schema in (13), showing nasal assimilation and nasal deletion, illustrated two forces

at work in creating the optimal output. However, using rewrite rules and derivations does

not reveal that two processes are working together in order to create the output; but

instead, it appears that two independent processes are working independent of one

another in order to create the output. A better model that illustrates that these processes

are interwoven and working together is applied below, using Optimality Theory, rather

than rules and processes that occur independently.

In this final chapter, the constraints on Brazilian Portuguese syllable phonotactics

will be ranked and that ranking will be investigated to determine whether or not the

native speaker of BP transfers their L1 ranking into their L2 English. First, a brief

introduction to Optimality Theory is given and the constraints used in the analysis are

introduced. Then a ranking for the syllable phonotactics of BP is introduced, and finally it

is compared to the inter-language of the BP speaker’s English.

28

6.1 OT: Introduction and Constraints 6.1.1 Introduction Optimality Theory (Prince and Smolensky 1993; Prince and McCarthy 1993) is a

relatively recent theoretical framework within generative phonology in which the surface

forms of language reflect resolutions of conflicts between competing constraints (Kager

1999). The ranking of constraints provides an understanding of the phonology of the

particular language. The higher the constraint is ranked the stronger it is, and the lower it

is ranked the weaker the constraint is in the language.8 Constraints are cross-linguistic

and universal, and the ranking is language specific.

6.2 The Constraints and Their Ranking in Brazilian Portuguese In this section, the constraints used in the analysis are introduced. A definition of each

constraint is given as well as its ranking in contrast with other constraints that play a role

in the syllable structure in BP. First, two faithfulness constraints are introduced. The

function of a faithfulness constraint is to ensure that the language retains segments in the

output that appear in the input, as well as disallowing new segments in the output that

were not in the input, essentially retaining the correspondence of segments that appear in

both the input and output forms.

(17) DEP-IO Output segments must have input correspondents (no epenthesis)

(McCarthy and Prince 1995)

8 For more information regarding Optimality Theory consult Prince and Smolensky 1993; Prince and McCarthy 1993a and 1993b; Archangeli 1997; and Kager 1999.

29

This constraint states that all segments in the output must have a corresponding

segment in the input, and that any segment in the output not in the input is a violation of

this constraint. The next constraint also deals with faithfulness between input and output.

(18) MAX-IO Input segments must have output correspondents (no deletion) (McCarthy

and Prince 1995)

MAX-IO states that all segments in the input must also be in the output form, and

any segment in the input that is deleted is a violation of this constraint. Here is an

example regarding the interaction between the two constraints.

Tab. 1 (gnomo ‘gnome’) /gnomo/ MAX-IO DEP-IO

Fa) gi.n�.@mu *

b) g�@.mu *!

In tableaux 1, the correct output form violates DEP-IO, but because it is lower

ranked and the other output candidate violates the higher ranked MAX-IO, (a) is the

optimal output. Now let’s use the same example and introduce a few syllable well-

formedness constraints. The function of well-formedness constraints are to enforce the

language’s permissible syllable constructions from the input forms into the output forms.

The next constraint deals with the onset of the syllable. The onset of the syllable

is the consonant or string of consonants that precede the nucleus or sonority peak

(typically a vowel) of the syllable (see §1.1 and §3.4 for the syllable structure of BP).

(19) *óó[Obs + [-Appx]: Obstruents plus non-approximants (obstruents and nasals) are

not permissible in onsets

30

This constraint states that the only consonant cluster permissible in the onset of

BP consists of an obstruent plus an approximant, and that obstruent plus nasal and

obstruent plus obstruent clusters are prohibited. This constraint is important for breaking

up impermissible consonant clusters such as *gn in gnomo.

Tab. 2 /gnomo/ *ó[Obs +

[-Appx] MAX-IO DEP-IO

a)gn�@.mu *!

Fb)gi.n�@.mu *

c)g�@.mu *!

In tableaux 2, *ó[Obs + [-Appx] and MAX-IO both dominate DEP-IO. This

ranking selects the optimal output form, which, through epenthesis, creates the

typologically ideal CV syllable. Epenthesis comes at the expense of DEP-IO, but because

it is low ranked, and it causes the satisfaction of *ó[Obs + [-Appx], the violation is

tolerated.

As was previously mentioned, BP prefers a simple unmarked CV syllable

construction, and the only consonants allowed in the coda are [s] or [r]. In the world’s

languages, the most unmarked syllable type is CV and the constraint in (20) retains the

ideal typological syllable structure by stating that syllables must be open, and coda

consonants are disallowed.

(20) NoCODA (*C]ó)ó) Syllables are open (Prince and Smolensky 1993)

We know that this is not always the case in BP, but BP does prefer the CV

syllable typology. The next constraint to be introduced also deals with the coda of the

31

syllable, and it states that consonant clusters are not allowed in the coda, or only one

consonant is permissible in the coda.

(21) *ComplexCODA (*CC]ó)ó) Codas must be simple (Prince and Smolensky 1993)

We would expect to find *ComplexCODA dominating NoCODA, because as was

mentioned earlier, BP allows the coda position of a syllable to be filled, even though it is

very restrictive with regards to which consonants are permissible in the coda position.

Here is the ranking of *ComplexCODA against NoCODA.

Tab. 3 (pesti ‘plague’)

/pE@sti/ *ComplexCODA NoCODA

Fa)pE@s.ti *

b)pE@st.i *! *

Tableaux 3 illustrates that BP allows certain consonants in the coda position but

no clusters (see §2.3). If /pE@sti/ ‘plague’ is syllabified so that the final vowel acts as its

own syllable by syllabifying all the preceding consonants in the coda of the preceding

syllable, as in (b), the phonotactics of the language are violated. Furthermore, this type of

construction would violate general conceptions of syllabification by not complying with

the unmarked CV syllable construction, because we have maximized the coda position of

the first syllable, while leaving the onset of the following syllable empty. Candidate (b)

above would lose regardless of the ranking, but complex codas are fixed and simple

codas are not because BP allows [s] and [r] in the coda. Now, using the previous

example, /pE@sti/ ‘plague’, we shall rank all five constraints against one another.

32

Tab. 4

/pE@sti/ MAX-IO *ó[Obs + [-Appx]

*Complex CODA

DEP-IO NoCODA

a) pE@.si.ti. *!

b) pE@.si *!

c) pE@.ti *!

d) pE@.sti *!

Fe) pE@s.ti *

f)pE@st.i *! *

In tableaux 1 it was determined that MAX-IO >> DEP-IO, and because the

ranking holds across the language, MAX-IO dominates DEP-IO in tableaux 4. Output

candidate (e) is optimal as it satisfies the higher ranked constraints. Output form (a), the

output form that fosters the ideal syllable typology, CV, loses because it violates DEP-IO,

which dominates NoCODA. Furthermore, the other output forms lose either because they

contain a non-permissible onset or coda cluster, or because a segment was deleted,

violating MAX-IO.

Next, another well-formedness constraint is introduced. All languages have onsets

and none disallow them. Not all languages allow onsetless syllables; however, BP does,

making this constraint low ranked in BP.

(22) ONSET (ONS) Onsets are obligatory (Prince and Smolensky 1993)

33

Tab. 5 (tabuada ‘multiplication table’)

/tabuada/ MAX-IO DEP-IO ONSET

a) ta.bu.ta @.dà *!

F b)ta.bu.a@.dà *

c) ta.bu.dà *!

In tableaux 5, both MAX-IO and DEP-IO outrank ONSET, as (b), the optimal

output candidate violates ONSET but satisfies both MAX-IO and DEP-IO, illustrating

that ONSET is low ranked. By epenthesizing a consonant to create a syllable onset,

satisfying ONSET, we violate DEP-IO, the higher ranked constraint. Finally, in (c), it is

noticed that if the onsetless syllable is deleted altogether, MAX-IO, the highest ranked

constraint, is violated.

Now we can address the topic of regressive nasalization and nasal consonant

deletion. As it was previously mentioned, BP does not allow nasal consonants in the coda

position of the syllable in the output. Therefore, the nasal consonant regressively

assimilates its nasalization onto the previous vowel and is deleted. However, this results

in the violation of markedness constraint:

(23) *Vnasal: Vowels must not be nasal (Kager 1999)

This constraint states that vowels are not to be nasal, regardless of whether or not

they are followed by a nasal consonant. The next constraint states that if a vowel precedes

a nasal consonant, then the vowel must be nasalized.

(24) *VoralN: Before a tautosyllabic nasal, vowels must not be oral (Kager 1999)

34

That is, a vowel must not remain oral when followed by a nasal consonant. The

nasal feature of the nasal consonant must regressively assimilate onto the previous vowel.

This constraint is important in English, where, when a nasal consonant is preceded by a

vowel, that vowel gains the nasal feature of the following nasal consonant, such as [sæ)nd]

‘sand’. The next constraint is another MAX-IO faithfulness constraint.

(25) MAX-IO (N): If the feature [+nasal] is in the input form then it must also be in the

output form. (McCarthy and Prince 1995)

The purpose of this constraint is for the feature [+nasal] to be retained in the

output form if it is present in the input form. The next constraint is another faithfulness

constraint that states that if a segment in the input is nasal then it must remain nasal in the

output form, and if a segment in the input form is non-nasal then it does not become nasal

in the output form.

(26) IDENT-IO (Nasal): Corresponding segments in the input and output must have

identical values for [nasal] (McCarthy and Prince 1995)

Here is the ranking for regressive assimilation of nasality onto the preceding

vowel and nasal consonant deletion.

Tab. 6 (transpor ‘to transpose, transport’) /transpor/ *Complex-

CODA MAX-IO (NASAL)

MAX-IO

IDENT-IO (NASAL)

*Vnasal

a) tra )ns.por *! * *

b) trans.por *!

c) tras.por *! *

Fd) tra )s.por * * *

35

Tableaux 6 displays the constraint ranking for regressive nasalization and nasal

consonant deletion. Suboptimal output candidates (a) and (b) lose on a violation of

*ComplexCODA because they retain the nasal consonant in the coda, creating a complex

coda in the output. Candidate (c) satisfies *ComplexCODA because of the deletion of [n]

but loses on MAX-IO (NASAL), as the feature [nasal] is lost from the input to the output.

Output (d) is optimal because it satisfies the two highest ranked constraints

*ComplexCODA and MAX-IO (NASAL). These constraints illustrate that in BP, the

nasal consonant must be deleted to avoid complex coda consonant clusters, and that the

feature [nasal], which is realized now on the vowel, must be retained in the output from

the input. The optimal output satisfies all these criteria, even though it violates, MAX-IO,

IDENT-IO (NASAL) and *Vnasal.

The next phonological process to be discussed is lateral gliding. Earlier, it was

discussed that, in BP, if a lateral occurs in the coda position it is glided.

(27) MAX (LATERAL): If the feature [+lateral] appears in the input form then it must

also appear in the output form (McCarthy and Prince 1995)

This constraint states that if a segment in the input has a positive value for the

feature [lateral] then it must retain that value in the output, thus lateral gliding will violate

this constraint.

The motivation behind lateral gliding cannot be understood using only the

constraints introduced so far. Lateral gliding occurs because [l] is not permissible on the

surface in the coda position and must be glided and syllabified into the nucleus. The

lateral is not deleted altogether, as that would incur a violation of MAX-IO, instead it

surfaces as a glide. Thus, a coda condition constraint must be introduced, because as it

36

stands now, with using only NoCODA and *ComplexCODA, there is no reason for [l] to

be forced into the syllable nucleus. Having an [l] in the coda position, is no worse than [s]

or [r] according to NoCODA by definition, and as long as [l] surfaces by itself, no

violation of *ComplexCODA is incurred. Thus, there is generalization that is being

missed when only these two constraints are used in conjunction. The generalization is

that BP allows only [s] and [r] in the coda position, and any other segment that occurs in

the coda underlyingly undergoes some phonological process to satisfy this coda

condition. It has been previously mentioned that BP allows only [s] and [r] in the syllable

coda, and thus a coda condition constraint stating such must be introduced.

(28) CODA-COND: Only [s] and [r] are permissible in the syllable coda

Tab. 7 (fa@cil ‘easy’)

/fa )sil/ CODA-COND

MAX-IO DEP-IO MAX-IO (LATERAL)

a) fa ).sil *!

Fb) fa ).siw *

c) fa @s.il *!

d) fa @.si *! *

e) fa @.si.li *!

Tableaux 7 shows the ranking for lateral gliding in BP. Suboptimal output

candidates (a) and (c) lose on a violation of CODA-COND, as the lateral [l] occurs in the

coda position, and candidate (d) loses on MAX-IO, as there is no corresponding segment

in the output for the input [l]. Candidate (e) loses on the account of DEP-IO being ranked

higher than MAX-IO (LATERAL). Optimal candidate (b) wins on the satisfaction of

37

MAX-IO, as the lateral surfaces as [w] and the segment is not lost, and there are no

consonants in the coda, satisfying CODA-COND.

What tableaux 7 illustrates is that BP wants to satisfy potential violations of

CODA-COND by gliding [l] to [w] and syllabifying it into the nucleus but not deleting it,

as that would incur a violation of MAX-IO. This is all done at the expense of low ranked

MAX-IO (L).

CODA-COND can also be implemented in the nasalization analysis, instead of

using *ComplexCODA. The advantage of CODA-COND over using NoCODA and

*ComplexCODA is not only that CODA-COND is a combination of both markedness

constraints, but also that CODA-COND captures the generalization missed by the two

constraints. In both nasalization and lateral gliding, a constraint, that of CODA-COND, is

forcing a simplification of the coda, either through nasal deletion, leaving only [s] in the

coda of /ns/ clusters, or lateral gliding, which syllabifies /l/ into the nucleus as [w],

satisfying CODA-COND. Furthermore, BP wants to retain some feature of the original

segment, either the nasalization realized on the vowel, or the glide that is syllabified in

the nucleus, which was [l] in the underlying representation.

6.3 Interlanguage Constraint Ranking This next section will compare the constraint ranking of the interlanguage English to both

English and BP, using the same constraints introduced in §6.2. The ranking of

faithfulness and markedness constraints with regards to their presence in the

interlanguage, English and BP, is discussed, as well as how the interlanguage constraint

ranking compared to both the target constraint ranking of English and the native

38

constraint ranking of BP. First, the nasalization process is discussed (§6.3.1), then lateral

gliding (§6.3.2) and finally epenthesis (§6.3.3).

6.3.1 Interlanguage Nasalization

In (§2.5) regressive assimilation of nasality and nasal deletion in BP was discussed. The

[nasal] feature of a nasal consonant spreads onto the preceding vowel and the nasal

consonant is deleted, in some cases to prevent a complex coda. Examples of this are

found in (4), which illustrated that this process only occurred when the nasal was in the

coda position.

In (§6.2) regressive assimilation of nasality and nasal deletion in BP was

explained using a constraint-based approach. The constraint ranking formulated in the

above analysis illustrated that there were two important processes occurring during the

assimilation and deletion. First was the elimination of the complex coda /ns/ by the high

ranking *ComplexCODA (or now CODA-COND), which forced the deletion of [n]. The

second was that even though the nasal [n] was deleted, its feature [nasal] must be retained

in the output, which surfaced on the vowel, motivated by MAX-IO (NASAL). These two

constraints were ranked against MAX-IO, IDENT-IO (NASAL) and *Vnasal. Tableaux 6

illustrated this ranking for the form transpor ‘to transpose, transport.’

In English, regressive assimilation of nasality also occurs, but the nasal consonant

is retained in the output. This was illustrated in (14), which introduced a contrastive

analysis of English and the interlanguage English collected in the data. A ranking for

English is introduced here and then followed by the interlanguage ranking. The tableaux

below illustrates that if *ComplexCODA is dominant in English, as it is in BP, then the

39

coda is forced to be reduced as it is in (c), which thus produces the incorrect optimal

output form. This is not the case in English as complex syllable margins are permissible.

Tab. 8 ‘plant’ (English output using BP ranking) /plant/ *Complex

CODA MAX-IO (NASAL)

IDENT-IO (NASAL)

*Vnasal

a. plant *!

b. Fpla)nt *! * *

c. Lpla)t * *

d. plat *!

The ranking here is similar to the BP ranking in tableaux 6. Tableaux 8 illustrates

that another constraint is present, working to retain [n] in the output and that

*ComplexCODA must be low ranked as suboptimal candidate (c), which according to the

ranking in tableaux 8, is just as optimal as the correct candidate (b). The ranking in

tableaux 8 effectively eliminates (8a), the input faithful candidate, because English wants

to regressively assimilate nasality, motivated by high ranking *VoralN, and (8d) loses on

the fact that English wants to retain the [nasal] feature in the output. The conflict between

(8b) and (8c) arises because the feature [nasal] remains in both output candidates, and the

only difference is the presence of [n] in candidate (b) and the absence of [n] in candidate

(c). This means that a constraint forcing input segments to remain in the output is present.

A constraint of this type was introduced earlier in (18), MAX-IO, which states that input

segments must have output correspondents. Introducing this constraint, with *VoralN,

high into the ranking above will effectively eliminate (8c) as a possible output, as [n] is

lost from the input to the output, and choose (b) as optimal.

40

Tab. 9 ‘plant (reranking)’ /plant/ *VoralN MAX-IO MAX-IO

(NASAL) IDENT-IO (NASAL)

*Vnasal *Complex CODA

a. plant *! *

b. Fpla)nt * * *

c. pla)t *! * *

d. plat *!

Thus far, the constraint ranking for regressive assimilation of nasality in English

has been introduced and compared to the ranking for the same process in BP. The

difference lies in the fact that BP wants to satisfy *ComplexCODA or CODA-COND and

thus deletes the nasal consonant with nasality assimilated so that it can surface. However,

in English, if *ComplexCODA were ranked in the tableaux it would be ranked low,

beneath MAX-IO, as English would rather have a complex syllable margin than lose a

segment in the output. The constraint *VoralN motivates the assimilation and MAX-IO

either retains the segment in English or *ComplexCODA or CODA-COND motivates the

nasal consonant deletion in BP. In BP MAX-IO was ranked below *ComplexCODA or

CODA-COND.

Next the interlanguage constraint ranking for nasalization is formulated and

compared to the rankings in English and BP. In (§5.2), it was discussed that the native

speakers of BP generally transfer this process into their interlanguage English. The forms

used in the analysis below are from (14).

In the interlanguage English, the native BP speakers regressively assimilated the

nasality and then deleted the nasal consonant. This is identical to the process in BP. What

41

this illustrates is that the markedness constraint, *ComplexCODA or CODA-COND, are

transferred in the interlanguage with a dominant ranking, forcing the deletion of the

nasal, while the constraint MAX-IO (N) motivates the assimilation. The ranking should

then be identical to the ranking in BP for regressive assimilation of nasality and nasal

deletion. The interlanguage output form of /plant/ is [pla)t]. This is illustrated through the

same ranking as in Tableaux 6.

Tab. 10 ‘plant (interlanguage form)’ /plant/ *Complex

CODA MAX-IO (NASAL)

MAX-IO IDENT-IO (NASAL)

*Vnasal

a. plant *!

b. pla)nt *! * *

c. F pla)t * * *

d. plat *! *

The English optimal output form (10b) loses on the violation of high ranking

*ComplexCODA, and candidate (10c) wins on the satisfaction of *ComplexCODA. The

optimal interlanguage output form violates MAX-IO, but MAX-IO is ranked low,

illustrating that in the interlanguage it is better to violate MAX-IO, while satisfying the

higher ranked *ComplexCODA. The two typologies are illustrated in (29).

(29) Typologies of BP, English and IL English BP: *ComplexCODA, CODA-COND >> MAX-IO English: MAX-IO >> *ComplexCODA

Figure (29) shows the distinct rankings between English and BP. In BP, the

markedness constraints, *ComplexCODA or CODA-COND, dominate the faithfulness

constraint MAX-IO. This illustrates that BP would rather syllabify according to the

phonotactics of the language than to remain faithful to the input form. In English,

42

according to the typology in (29), it is more optimal to remain faithful to the input than to

satisfy certain markedness constraints.

The discussion thus far proposes that the process of regressive assimilation of

nasality and nasal deletion in BP is transferred into the in interlanguage along with the

native BP constraint ranking. This deviates from native English in the sense that the

faithfulness constraint MAX-IO is violated in the interlanguage.

6.3.2 Emergence of the Inactive Constraint: MAX-IO (OBS)

There are two important components omitted in the above section (§6.3.1). Firstly, a

possible output candidate has not been evaluated by the above ranking in Tableaux 10,

and secondly, a typology, that of the interlanguage English, has been omitted in (29).

The unevaluated candidate absent from Tableaux 10 is that of [pla)], which

according to the ranking, if CODA-COND was transferred into the interlanguage

phonology, as it should be, would prove to be more optimal than the optimal candidate

[pla)t], Tab. 10c. Thus, a constraint that ensures the retention of [t] in the output but does

not effect the loss of [n] must emerge into the ranking. This constraint could most aptly

be formulated as MAX-IO (OBS), which is introduced and defined below:

(30) MAX-IO (OBS): Every obstruent (oral stop, fricative, affricate) in the input

must be retained in the output.

The emergence of this constraint in the interlanguage retains the [t] and has no

effect on the retention or loss of [n]. The interaction of this constraint with

ComplexCODA and CODA-COND, along with MAX-IO (NASAL), which retains the

43

feature [nasal] from the input to the output, selects the optimal candidate exhibiting a

simplification of a complex coda, while retaining the feature [nasal] and the segment [t].

Tab. 11 ‘plant (refined interlanguage form)’ /plant/ MAX-IO

(OBS) *Complex

CODA MAX-IO (NASAL)

CODA-COND

MAX-IO IDENT-IO(NASAL)

*Vnasal

a. plant *! *

b. pla)nt *! * * *

c. F pla)t * * * *

d. plat *! * *

e. pla) *! * * *

The emergence of MAX-IO (OBS) in the interlanguage eliminates the otherwise

optimal candidate (e). Because the only consonants that occur in the coda position of the

input in BP are /s r n l/, the retention of [t] in the coda is never attested. This follows

Broselow (1998), whereby the interlanguage output does not necessarily reflect the

phonological system of the native language. Outputs exhibiting non-native like forms are

often the result of basic faithfulness constraints acting upon input forms divergent of

input forms present in the native language. Thus, unmarked outputs, with features

attributable to both the native and target languages, surface.

The typology for the interlanguage ranking differs from the BP ranking with the

emergence of MAX-IO (OBS). Inputs present in the target language that do not appear in

the native language often cause constraints, otherwise inactive in the native language, to

become active in the interlanguage. Constraints in OT do not operate on a binary system

like distinctive features do, rather all constraints are present in an OT grammar, and with

those higher-ranked having a greater influence on the output selection.

44

6.3.3 Lateral Gliding in the Coda Position

In (§3.4) lateral gliding in BP was discussed, and it illustrated that when a lateral, namely

[l], occurred in the coda position, it was glided and syllabified into the nucleus of the

syllable. The force behind lateral gliding is the high ranking of CODA-COND in BP.

The constraint ranking motivating this process was presented in tableaux 7, where

CODA-COND, primarily, forces the lateral to glide to [w] and be syllabified in the

nucleus. This alternation occurs at the expense of a violation of MAX-IO (LATERAL).

This process, which occurs in BP, does not occur in English, where it is better to

retain the feature [lateral] in the output from the input than to incur a violation of CODA-

COND. This is plausible in the sense that English is much less restrictive in the coda than

BP. In English, MAX -IO (LATERAL) would dominate CODA-COND. This is evident

in tableaux 12, which ranks the two constraints against one another using the form [t�l]

‘tall.’

Tab. 12 ‘tall’

/t�l/ MAX -IO (L) CODA-COND

a. F t�l *

b. t�w *!

Tableaux 12 shows the strict domination of MAX -IO (LATERAL) over CODA-

COND. Thus, in English it is better to have a segment occupying the coda position than

to lose the feature [lateral] in the output. In BP, the opposite was the case where CODA-

COND dominated MAX -IO (L) forcing /l/ to glide and syllabify into the coda, satisfying

CODA-COND.

45

In (§5.3) it was illustrated that this process, generally, is transferred into the

interlanguage English. This would mean that the ranking from BP holds in the

interlanguage, as opposed to the ranking in English. Thus, CODA-COND must dominate

MAX -IO (L). This is illustrated in tableaux 12.

Tab. 13 ‘tall (interlanguage form)’

/t�l/ CODA-COND

MAX -IO (L)

a. t�l *!

b. F t�w *

Tableaux 13 shows that when CODA-COND dominates MAX -IO (L) it is better

to glide the lateral than to violate CODA-COND. This ranking holds in both BP and the

interlanguage English, and the opposite ranking holds in native English. The two

typologies are present in (31).

(31) Lateral Gliding Typologies BP: CODA-COND >> MAX -IO (L) English: MAX -IO (L) >> CODA-COND IL English: CODA-COND >> MAX -IO (L) 6.3.4 Epenthesis in Interlanguage English The last interlanguage process to be discussed is epenthesis to dissolve possible

problematic complex syllable margins. Two types of epenthesis are possible, the first

entails physically breaking up a non-native consonant cluster. For example, if a

tautosyllabic [sÄ] cluster occurred underlyingly in some language X, and that language

did not permit such a cluster, then [«] could be epenthesized as such, [s«Ä], which breaks

up the cluster. The second type is to force a resyllabification of the adjacent segments in

violation by epenthesizing before the cluster. For example, taking the same cluster [sÄ]

46

and epenthesizing [«] before the [s] creating [«sÄ] where the [s] is syllabified into the

coda of the new syllable and [Ä] remains as the onset of the input syllable. This type of

epenthesis is present in Spanish, which has CONTIGUITY-IO, a constraint barring

against medial epenthesis or deletion of segments, ranked high.

In BP (§2.1) when words are borrowed that contain impermissible consonant

clusters, BP epenthesizes [i] to break up this cluster. This process was evaluated by

constraints in Optimality Theory in tableaus 1 and 2. In these rankings, DEP-IO ranked

lowest and was dominated by *ó[Obs + [-Appx].

In English complex syllable margins are permissible and thereby DEP-IO and

MAX-IO dominate *Complex. This is opposite of BP. In the interlanguage, as discussed

in (§5.1), it would be expected that the ranking of *Complex would be high, as it was in

the interlanguage account of nasalization, dominating MAX and DEP constraints, as BP

would rather satisfy *Complex and violate MAX and DEP constraints. So, in the

interlanguage English, this ranking should be transferred. In English, the MAX and DEP

constraints are ranked high, and in terms of breaking up complex syllable margins, it is

better to have a complex margin than violate the faithfulness constraints. The two

typologies are introduced in (32).

(32) Typologies for Epenthesis BP: *Complex >> MAX-IO, DEP-IO English: MAX-IO, DEP-IO >> *Complex

The only time English epenthesizes is in borrowed words, which is the same as in

BP. So in both cases the languages are very faithful to their inputs. The question in this

section is, however, if the native BP speakers would epenthesize when confronted with

non-BP clusters.

47

Also discussed in (§5.1) is that there was very little epenthesis occurring in the

collected data, at least not enough to make a generalization. Possible explanations for this

are explained in the conclusion. But to use the input faithful forms as optimal in the

interlanguage would comply with the ranking in English but violate the ranking in BP.

This counters what was found in (§6.3.1) and (§6.3.2), where the interlanguage anking

was transferred from BP, even though it went against the target English ranking.

7.0 Conclusion This paper has given a basic introduction to the contrasting phonologies of English and

BP, and more importantly discussed the transferring of constraint rankings from a native

language to an interlanguage. The BP processes of regressive assimilation of nasality and

nasal deletion, lateral gliding in the coda position and epenthesization were discussed,

and the rankings motivating these processes, using an Optimality Theoretic approach,

were formulated and analyzed. The interlanguage data was collected, analyzed and a

ranking for the interlanguage was formulated and then compared to the rankings in BP

and English.

Broselow, et. all (1998) argued that second language learners often have a

tendency to favor less marked structures in the interlanguage. Particular markedness

constraints that are not visible in the native language either surface because of the

richness of the target-language input or the development of a phonology that differs from

the native language phonology. In terms of this argument applied here, it is evident that

for one, BP already has a fairly unmarked syllable structure that motivates certain

phonological alternations, and from the data above, it is evident that the rankings were

48

transferred with regards to nasality and lateral gliding. However, there were isolated

cases of a more advanced second language learner of English that did not delete the nasal

and uttered the complex coda. This may be an example of the richness of the input forms

providing an alternative phonology from the native language. For example, BP does not

allow the fairly unmarked [t] in the coda position, as English does, but it is more likely

for that segment in that position to be produced in the coda in the interlanguage than

deleted or vowel epenthesis to syllabify it into the onset. This is evident in Broselow’s

argument that a less marked interlanguage is favored even though it may differ from the

native language. But this is left for future research.

In the cases of regressive nasality and lateral gliding, the rankings were

transferred from the native language to the interlanguage according to the acoustic data,

with an emergence of MAX-IO (OBS) in the interlanguage. In the case of epenthesis, the

expected results were not found and near native-like production was present. There are

two possible explanations for this. Either one, complex English onset and coda consonant

clusters are easier to produce for native speakers of BP, or, more likely, there was a fault

in the data collection. By placing the target forms between two vowels it may have eased

the pronunciation, because the vowels could act as the nucleus of a syllable and the

exterior most consonants in the clusters could have been syllabified around these vowels.

In order to fully validate the analysis, future research with more participants is needed as

well as a different method for eliciting the target words in question. However, because of

the strong evidence of transference in the limited data of nasalization and lateral gliding,

it is evident that the native language ranking is transferred into the interlanguage of the

target language.

49

8.0 References

Archangeli, Diana. 1997. Optimality Theory: An Introduction to Linguistics in the 1990s.

In Optimality Theory: An Overview, ed. Diana Archangeli and D. Terence

Langendoen, 1-32. Malden, Mass.: Blackwell Publishers.

Azevedo, Milton. 1981. A Contrastive Phonology of Portuguese and English.

Washington, D.C.: Georgetown University Press.

Broselow, Ellen, Su-I Chen and Chilin Wang. 1998. The Emergence of the Unmarked in

Second Language Phonology. Studies in Second Language Acquisition 20: 261-

280.

Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New York:

Harper and Row.

Gass, Susan and Larry Selinker. 1994. Second Language Acquisition: An Introductory

Course. Hillsdale, NJ: Lawrence Erlbaum Associates.

Giangola, James. 1997. Constraint Interaction and Brazilian Portuguese Glide

Distribution. Rutgers Optimality Archive (ROA 182-03973).

Girelli, Carl. 1988. Brazilian Portuguese Syllable Structure. Ph.D. Dissertation,

University of Connecticut.

Harris, James. 1974. Evidence from Portuguese for the ‘Elsewhere Condition’ in

Phonology. Linguistic Inquiry 5, 61-80.

Kager, Rene @. 1999. Optimality Theory. Cambridge: Cambridge University Press.

Ladefoged, Peter and Ian Maddieson. 1996. The Sounds of the World’s Languages.

Malden, MA: Blackwell.

Major, Roy. 1985. Stress and Rhythm in Brazilian Portuguese. Language 61:2; 259-282.

50

Mateus, Maria Helena and Ernesto d’Andrade. 2000. The Phonology of Portuguese.

Oxford, England: Oxford University Press.

McCarthy, John and Alan Prince. 1993a. Generalized Alignment. Technical Report #7 of

the Rutgers Center for Cognitive Science, Rutgers, NJ.

---------------. 1993b. Prosodic Morphology. Technical Report #3 of the Rutgers Center

for Cognitive Science, Rutgers, NJ.

---------------. 1995. ‘Faithfulness and Reproductive Identity’ in Beckman, Dickey, and

Urbanczyk (eds.), Papers in Optimality Theory (University of Massachusetts

Occasional Papers 18) Amherst: GLSA: 249-384.

Morales-Front, Alfonso and D. Eric Holt. 1997. On the Interplay of Morphology,

Prosody and Faithfulness in Portuguese Pluralization. In Issues in the Phonology

and Morphology of the Major Iberian Languages, ed. Fernando Martinez-Gil and

Alfonso Morales-Front, 391-437. Washington, D.C.: Georgetown University

Press.

Prince, Alan and Paul Smolensky. 1993. Optimality Theory: Constraint Ranking in

Generative Grammar. Tech. Rep. #2 of the Rutgers Center for Cognitive

Sciences, Rutgers, NJ.

Redenbarger, Wayne J. 1981. Articulator Features and Portuguese Vowel Height.

Department of Romance Languages, Harvard University, Cambridge, MA.

Reider, Michael. 1981. Theoretical Aspects of Denasalization in Brazilian

Portuguese. GLOSSA 15.2.

Spencer, Andrew. 1996. Phonology. Malden, MA: Blackwell.

51

Wetzels, W L. 1997. The Lexical Representation of Nasality in Brazilian Portuguese.

Probus 9, 203-232.